Evaluation of Sampling Techniques
on Tall-Grass Prairie
DONALD A. BECKER AND JERRY J. CROCKETT
Highlight: An evaluation of sampling techniques was conducted on a tall-grass
prairie in eastern 0 klahoma. The point transect, a modified point transect, line
transect, angle order, quarter, quadrat, and wandering quarter methods were used.
Relative and total density values were determined and compared with actual values
obtained by hand count. The above methods, with the exception of the quadrat,
underestimated the relative density of splitbeard bluestem (Andropogon
ternarius), a
dominant
and densely cloned species, and overestimated
switchgrass (Panicum
virgatum), a subdominant, single-s talked species. Relative and total density values
obtained by the quarter method were significantly less accurate than those obtained
by the other methods; no method was significantly more accurate. With recalculation
excluding splitbeard bluestem data, relative densities obtained by most of the methods
agreed more closely with actual values, and the quarter method was again significantly
less accurate. Results indicated that degree of clone density of the dominants and subdominants, as well as sampling time, should be noted prior to the selection of a sampling method in a highly aggregated grassland type. The modified point transect or
quadrat methods are considered to be most applicable if the dominants are densely
cloned as in splitbeard bluestem; however, the point transect or line transect methods
may be adequate if the dominants are sparsely-cloned as in big bluestem (Andropogon
gerardi) or little bluestem (Andropogon scoparius).
The quadrat method is one of the
oldest and most widely used methods for
grassland analysis. Other methods, however, have been developed to avoid much
of the sampling time needed to obtain
statistically reliable results. These methods include
the meter-line
transect
(Weaver and Clements,
1929); point
transect (Levy and Madden, 1933); loop
(Parker, 195 1); and quarter (Dix, 1961).
Whitman and Siggeirsson (1954) compared the line and point transect methods
in a mixed grassland type in North
Dakota and considered both to be equally
reproducible. Johnston (1957) compared
the point transect, line transect, and loop
methods in a mixed grassland type in
Alberta. The loop method was found to
be least reproducible
and a preference
was indicated
for the point transect
The authors are assistant professor of biology in the Department
of Biology, Midland
Lutheran
College,
Fremont,
Nebraska,
and
associate
professor
of plant ecology
in the
Department
of Botany and Plant Pathology,
Oklahoma State University, Stillwater.
This work was done while the senior author
was a participant in a NSF Summer Research
Participation
Program in botany at Oklahoma
State University during the summer of 1970.
The authors wish to acknowledge Dr. Gerald
M. Funk for assistance with statistical procedures.
Manuscript received December 3 1, 197 1.
JOURNAL
OF
RANGE
MANAGEMENT
26(l),
method. Crockett (1963) compared the
line transect, point transect, and quarter
methods in a tall grassland type in Oklahoma and indicated a preference for the
point transect method. More recently,
density data obtained by the quarter
method was compared to quadrat data
for tall grassland types in Wisconsin
and in
(Risser
and Zedler , 1968)
Minnesota (Good and Good, 1971). In
these studies the quarter method results
did not closely agree with those of the
quadrat method, particularly
in aggregated stands.
The above studies were somewhat
limited in scope in that only a few
methods were compared in a stand, and
sampled and actual counts were not made
to determine accuracy. Nor were methods
for non-random
populations,
the angle
order (Morisita,
1957) and wandering
quarter (Catana, 1963) compared with
those best adapted for random populatior s (quarter, line transect, quadrat).
The objectives of this study were to
the field adaptability
of
compare
methods in a highly aggregated tall-grass
prairie; to determine accuracy by comparing actual and sampled data for the
various methods; and to determine if
accuracy of methods for non-random
January
1973
populations
was greater
random populations.
than those for
Methods
The study site was located in an
upland prairie, approximately
11 km west
of Stillwater, Okla. By visual inspection
the stand appeared homogeneous
and
contained as a dominant splitbeard bluestem (Andropogon ternarius), and as subswitchgrass
(Panicum
virdominants
gatum), little bluestem (A. scoparius), big
bluestem (A. gerardi), and broomsedge
(A. virginicus). The vegetation was approximately 30-60 cm high when sampled
in July of 1970, and a 4-8 cm mulch layer
was present. Four plots, each 1.2 x 5 m,
established
and
were
systematically
numbered l-4. Each was subdivided at 20
cm intervals by imaginary coordinates
projected from stakes on the plot margins. Four sampling methods were used
on these plots; the point transect (10
pins), the meter-line transect, the quarter,
and the angle order. On plots 3 and 4
three other methods were used: a modification of the point transect, whereby all
basal contacts made by each pin were
counted;
a 10 x 10 cm open-ended
quadrat; and the wandering quarter. Relative density (percentage composition) for
each major species was determined by all
of the above methods, relative frequency
(frequency value of a species divided by
total of frequency values for all species)
by all except the wandering quarter, and
total density by all except the point
transect and wandering quarter methods.
The sampling units were randomized
on each of the plots by the selection of a
series of numbers, each of which represented
a particular
coordinate
intersection. Sampling on each plot was terminated when approximately 400 shoots
were encountered.
A bit or contact by
the point transect methods was determined when a shoot was hit within 0.5
cm of ground level, and by the line
transect method when a shoot was found
within a 0.5 cm strip on either side of a 1
mm cord stretched tightly between two
pins. After sampling,
vegetation
was
removed from the plot and all shoots
counted to determine the actual relative
61
and total density.
Determination
of total density in the
quarter and angle order methods followed
Dix (1961) and Morisita (1957), respectively. In the latter method two estimates
are determined,
m r and m2. If ml is
X obs = the number
X exp =
-
than iiiz, ?
greater
is considered
as the
best density estimate;
if ii-r, is less than
iii2, the average of the two estimates
m.
m 0 is considered
as
is determined,
and
Ir
the best estimate of density.
Since both sampled
and actual data
were collected,
the sampled
and actual
ratios of the populations
could be compared by the chi-square statistic given by
the formula:
x2 =
where
c
(xobs
- xe~p)~
GXP
x2 = the chi-square
c
value
= the
summation
of the ratios
obtained when observed and actual data
for each species or group of species
(i.e., forbs and miscellaneous
grasses)
were compared.
Since there were
seven species or groups of species (including splitbeard
bluestem)
and six
species or groups of species excluding it, seven or six ratios were summed
to obtain
the respective
chi-square
value.
Table 1.
A comparison
Species
Splitbeard
bluestem
Switchgrass
Little bluestem
Big bluestem
Broomsedge
Miscellaneous
Forbs
bluestem
grasses
the sum total of shoots of all species
encountered
in that plot multiplied
by the actual relative density of that
species.
The chi-square values for each method
on all four plots were compared
for
significance
by
nonparametric
procedures: the Friedman
statistic (Bradley,
1968),
or the Kruskall-Wallis
statistic
(Siegel, 1956).
Two indices of aggregation were calculated to determine
vegetation
pattern.
These
included
the
index
of nonrandomness
(Pielou,
1959) and the variance: mean ratio (Blackman,
1942). The
former is determined
from the relation
ac = 71D cc), where c3 is the mean square
of the point to plant distances
of the
quarter method
and D is the density of
plants determined
by actual counts. The
latter index was calculated from quadrat
data, and could be determined
on plots
3 and 4 only. In both indices, values
in excess of 1 .O indicate aggregation.
Results
Density
Sampled values of relative density did
not closely approximate actual values for
the major species by the methods used
(Table 1). 0 n all plots most of the
of sampled and actual density values (%) obtained
by the sampling methods.’
Plot
Actual
relative
density
Point
transect
Line
transect
Angle2
order
Quarter
Modified
point
transect
1
2
62.7
51.9
46.8
40.8
56.3
40.1
52.8
47.0
43.8
37.8
-
3
4
51.9
37.5
30.2
26.3
36.9
30.0
29.6
31.5
20.3
22.0
38.9
32.1
45.3
29.4
36.6
34.2
1
2
10.1
16.5
19.9
19.6
19.5
23.2
18.5
24.8
31.3
31.8
-
_
-
-
3
4
19.0
15.7
24.0
19.0
27.0
17.4
34.6
24.8
43.3
39.0
20.9
20.3
18.3
15.3
31.8
17.6
1
2
8.0
17.4
8.7
17.9
6.0
17.4
8.5
13.8
6.8
11.0
-
-
-
3
4
10.5
21.3
19.1
25.1
18.4
25.6
12.3
20.0
8.3
12.3
17.3
26.7
19.7
28.5
14.5
14.5
1
2
13.3
4.1
15.9
4.1
11.8
7.5
12.0
2.5
9.5
2.0
-
_
3
4
7.6
15.4
11.4
16.1
7.6
16.4
6.3
11.5
4.8
8.0
10.6
16.1
4.6
14.6
6.1
24.5
1
2
0.8
2.2
1.5
3.1
0.2
0.7
1.3
2.3
1.5
1.5
-
-
-
3
4
4.0
2.0
6.2
4.6
1.8
1.7
5.8
3.8
3.8
5.3
6.0
2.0
3.9
1.5
3.8
1.5
1
2
3.5
5.3
3.8
12.2
4.3
7.5
4.8
7.0
5.3
10.0
-
-
-
3
4
4.4
5.2
6.4
5.1
5.6
5.2
5.3
4.5
12.3
5.5
4.8
1.5
2.7
6.3
4.1
4.6
1
2
1.8
2.7
3.3
2.3
2.4
3.5
2.3
2.3
2.0
6.3
-
-
-
3
4
2.6
2.9
Z:::
2.7
3.7
6.3
4.0
7.5
8.0
1.7
1.5
5.6
4.3
3.1
3.1
‘The quadrat, modified point transect,
2n=3; k=4 (see Morisita, 1957).
62
methods, but particularly
the quarter,
underestimated
the relative density of
splitbeard
bluestem,
a densely-cloned
species, and overestimated switchgrass, a
single-stalked species. The latter was best
estimated
by the quadrat
method.
Loosely-cloned species such as little bluestem and big bluestem
were underestimated by the quarter method but not
by the point transect, line transect, or
angle order methods. Other species or
groups such as broomsedge
bluestem,
miscellaneous grasses, and forbs were not
at all or only slightly underestimated
by
the point transect, line transect, and angle
order techniques; however, the quarter
method grossly overestimated the singlestalked miscellaneous grasses and forbs.
Chi-square values calculated for the point
transect, line transect, and angle order
methods were approximately
equivalent,
while those of the quarter method were
much higher (Table 2). Only the chisquare value of 11.59 on plot 4 with the
line transect method indicated a good fit
between actual and sampled data (P =
.05). When the chi-square values were
ranked and compared by the Friedman
test, however, only the values obtained
by the quarter method were significantly
less accurate (P = .05). Splitbeard blue-
of shoots of a species on
a particular
plot encountered
by the
sampling method.
and wandering
quarter
methods
Quadrat
Wandering
quarter
-
-
were used on plots 3-4 only.
JOURNAL
OF
RANGE
MANAGEMENT
26(l),
January
1973
Table 2.
method
Chi-square
values determined
from actual and sampled density
on each plot are given as well as average and overall va1ues.l
Point
transect
Line
transect
63.3
52.9
86.6
34.8
58.2
166.1
45.0
44.5
67.7
11.6
42.2
110.3
1
2
15.7
23.7
31.5
16.7
3
4
F
Ox2
7.1
11.0
14.4
9.2
20.6
1.7
17.6
27.9
Plot
1
2
3
4
F
Ox2
Angle
order
Modified
point
transect
Quarter
Including
38.1
splitbeard bluestem
211.3
26.6
115.8
37.2
54.6
175.1
73.5
299.7
250.8
764.6
206.5
Excluding
17.4
20.9
splitbeard bluestem
100.9
68.5
24.6
28.3
22.8
78.1
83.8
168.0
105.3
349.4
‘The overall chi-square value for plots 14 or 3-4 is indicated
dense clumps,
stem, a species with
appeared to contribute to the poor agreement of sampled and actual data, and
data were recalculated with its exclusion.
Subsequently,
the recalculated data for
the major species (switchgrass, little bluestem, big bluestem) coincided to a greater
degree with actual values, particularly for
the point and line transect methods, but
not as markedly for the quarter method.
Chi-square values from the point transect,
line transect, and angle order methods
tended to be considerably
lower than
those of the quarter method (Table 2).
Ranking and comparing the chi-square
values from plots l-4 with the Friedman
statistic,
again indicated
the quarter
method was significantly
less accurate,
but there were no significant differences
between the angle order, point transect,
and line transect methods.
On plots 3 and 4 somewhat better
agreement between actual and sampled
data for three of the major species
(switchgrass, splitbeard bluestem, and big
bluestem) was obtained by the quadrat
and modified point transect methods, but
not for the wandering ,?uarter method. In
fact, lower average at,d overall chi-square
values for plots 3 and 4 were obtained by
the modified point transect and quadrat
methods
than by all other methods
(Table 2). To enable statistical
comparison between all methods (except the
quarter method) on plots 3-4 only, the
results were diiided into 10 subgroups,
five for each plot. With subsequent recalculation of chi-square values, ranking, and
application of the Friedman statistic, no
significant difference was found (P<O.20).
Exclusion of splitbeard bluestem
data
also resulted in a better fit of sampled
and actual data for the major species by
the three above methods.
Lower chiJOURNAL
OF
RANGE
MANAGEMENT
26(l),
data.
Values
the line transect method, but not significantly more accurate than that of the
quadrat method (P = .OS). The quadrat
method was not significantly
more accurate than the angle order or line transect method.
for each
Wandering
quarter
Quadrat
-
-
-
43.3
27.7
56.2
35.5
57.3
20.7
63.7
39.0
59.3
32.3
45.8
52.9
-
-
-
11.6
20.9
16.2
22.0
44.4
8.6
26.5
39.3
F7.1
29.0
23.1
19.8
Frequency
Relative frequency varied considerably
with the different sampling methods. The
point transect and angle order methods
ranked splitbeard
bluestem
first and
switchgrass second in frequency; the line
transect and quarter methods ranked
splitbeard bluestem and switchgrass about
equal in frequency;
and the quadrat
method ranked switchgrass first and splitbeard bluestem second in frequency. In
all methods little bluestem and big bluestem ranked as the third and fourth most
frequent species, respectively. A comparison of density and frequency data indicated that for splitbeard bluestem, the
most dense species, higher density values
relative to freqaency were obtained by
the quadrat and line transect methods.
Conversely, lower density values relative
to frequency were obtained for singlestalked species such as switchgrass, miscellaneous grasses, and forbs. Frequency
values determined by the point transect,
angle order, and quarter methods were
approximately
of the same order of
magnitude
as their respective density
values. These data indicated that the
frequency
values determined
by the
quadrat and line transect methods best
reflected
the dispersion of the major
species or species groups.
Pielou’s and the variance:mean indices
indicated that the vegetation was aggregated on all plots (Table 4). Those indices
calculated by Pielou’s index were all high
(7.96-l 1.05) as were those calculated by
the variance:mean
index on plot 3
(10.10) and plot 4 (4.39). A concomitant
by 0x2.
square values were also obtained by the
modified
point transect
method,
but
these again were not significantly more
accurate.
Actual and sampled total density values are given in Table 3. To enable
statistical comparison, each of the plot
results was divided into five subgroups.
Chi-square values were then calculated,
ranked, and compared by the KruskallWallis test. Density estimates obtained by
the quarter method were significantly less
accurate (P = .OOl). Sampled values from
the quadrat method were closest to actual
values unless the 28 measurements
1 cm
or less were excluded
from the !!!! densi-7r
ty estimate of the angle order method
(Table 3). When the sum of reciprocals of
these numbers was included, the estimate
!?!! was disproportionately
increased and
71
density was greatly overestimated.
The
estimate excluding the 28 measurements,
iii
lex,
was significantly
more accurate
lr
by !!!!, !%, or by
71 lr
than those obtained
Table 3.
Comparison
Plot
Actual
density
of total density
values (plants/m2)
obtained
from some sampling
techniques.
Angle order
Line
transect
Quarter
Quadrat
*_
mlex
**_
ml
n
1
2
3
4
X=
1000.5
1073.4
983.3
889.7
986.7
*iiilex
was calculated
71
**iii1 was calculated
836.0
870.0
1310.0
1382.0
1099.5
372.2
404.0
535.6
376.1
422.0
by excluding
by including
1056.0
962.0
1009.0
distance
all distance
measurements
1000.5
952.6
916.6
819.8
923.6
7
1897.1
1143.6
1362.5
1966.3
1592.4
***_
m0
71
1321.3
877.2
980.1
1350.1
1132.2
one cm or less.
measurements.
-F
***iii,
is the average of ml and m2 divided
by 71.
Ir
January
1973
63
decrease in the aggregation indices and in
relative density of splitbeard bluestem,
the most aggregated species, is also indicated in Table 4.
The methods varied considerably
in
sampling time and could be ranked according to decreasing time requirements
as follows: point transect, line transect,
angle order, wandering quarter, quarter
and quadrat. Approximately
400 shoots
could be encountered
by the quadrat
method in one-half less time than in the
point transect method. Sampling time for
both the point and line transect methods
was increased because of the mulch layer;
in addition, the high density of splitbeard
clumps
(5-6 shoots/cm2)
hampered
sampling in the latter. In the angle order
method much time was spent in locating
the third nearest individual and calculating total density.
Discussion
The composition
of an aggregated
grassland community
theoretically
can
best be determined
by a single pin of
infinitely small size as a sampling unit,
which is repeatedly
lowered into the
vegetation,
striking the shoots of each
species. In a random sample the probability of hitting shoots of each species is
proportional
to their density. Since only
sampling error would be present, the
observed values should closely approximate the expected values. The chi-square
values obtained by such a method should
not exceed 12.59, with six degrees of
freedom
at P = .05 (Snedecor and
Cochran, 1967). The poor agreement of
sampled and actual data in this study
indicates considerable
bias in the sampling methods. High aggregation of some
species is believed to contribute to much
of the poor agreement. Splitbeard bluestem was an especially aggregated species;
a decrease in its density coincided with a
decrease in magnitude of Pielou’s index
on plots 1-4. In addition, sampled values
departed less from actual values for major
species when splitbeard bluestem data
were excluded. With its inclusion the
chi-square value of 12.59 was not or was
slightly exceeded only by the modified
point transect method on plot 3 and 4,
and by the line transect method on plot
4. With its exclusion, however, the maximum expected chi-square value of 11.07
(5 degrees of freedom at P = .05) was
obtained on plot 3 by both of the point
transect methods, and on plot 4 by the
point transect, line transect, and quadrat
methods.
Density data obtained by the quarter
method were significantly less accurate
64
Table 4. Two measures of aggregation.
at P = .Ol.
Plot
Pielou’s index
1
2
3
4
11.05
9.97
8.90
7.96
Variance: mean1
upon whether
62.7
51.9
51.9
37.5
only for plots 3 and 4.
than those obtained by the other methods. The quarter method was originally
developed
to sample dispersed forest
vegetation, but subsequently was used in
grasslands by Dix (1961). The reliability
of this method for grasslands was evaluated in several studies in which quarter
method density data were compared to
either actual or quadrat counts. In the
Arizona desert grassland and California
grassland types, Morris (1962) found the
quarter method grossly underestimated
density. In Wisconsin, Risser and Zedler
(1968) found that density was grossly
underestimated
for stands dominated by
aggregated species such as little bluestem,
big bluestem,
prairie dropseed (Sporobolus
heterolepis),
or hairy
grama
(Bouteloua hirsuta). In Minnesota, Good
and Good (1971) found that the quarter
method underestimated
density in xeric
dominated
by grama grass
stands
(Bou teloua gracilis), needleandthread
(Stipc! comata), and in mesic stands dominated by big bluestem and Kentucky
bluegrass (Poa pratensis).
In this study better results were obtained by distance methods other than
the quarter method; for example, the
wandering quarter for relative density,
and the angle order for total density. The
accuracy of the estimates of the latter,
depended
Rel. density of splitbeard bluestem (%)
10.10
4.39
‘Quadrat data were obtained
however,
significant
mlex,
lr
plots 3 and 4, when compared with the
other methods, were not significantly
different.
One of these, the modified
point transect method, does appear to
have less bias in this highly aggregated
grassland type because, on the average,
chi-square values calculated from the 10
subgroups of relative density data were
lower. When splitbeard bluestem data
were omitted,
however, the modified
point transect did not appear to have any
advantage over the line transect and point
transect methods. These latter methods
may give accurate results when used to
sample species with loose clones such as
big bluestem and little bluestem.
The choice of a particular method for
sampling grassland vegetation should be
based not only on growth form, but also
on the degree of aggregation of the
dominant
and subdominant
species.
Accuracy, as well as time and reproduceability, should also be considered. In
this grassland type the quadrat method
appeared to have advantages over the
other methods when time was considered,
but did not appear to be as accurate as
the point transect method. The development of better techniques to assess density and percentage composition of highly
aggregated species is needed to facilitate
studies in aggregated grassland communities.
%_!
, or
!% was used. In this study the
7T
n
estimate iii1 was greater than iii2 in all
plots, and according to Morisita (1957),
?!?! should be used as the density estin
iii
However, only lex
gave a good
Yr
density estimate; and it was significantly
ii-l or the line
more accurate than iii1
-,
O,
ll
n
transect method. In a desert grassland
type in Arizona, Morris (1962) compared
angle order density data with that obtained by quadrat and actual counts. In
that study, despite the fact that i?il was
mate.
greater than iii:!, !!& rather than ml
lr
iF gave
the better density estimate.
The three additional methods used on
JOURNAL
Literature Cited
Blackman, G. E. 1942. Statistical and ecological
studies in the distribution of species in plant
communities. 1. Dispersion as a factor in the
study of changes in plant populations.
Ann. of Bot. 6:35 l-370.
Bradley, J. E. 1968. Distribution-free statistical
tests Prentice-Hall, Inc., Englewood Cliffs,
N. J. 338 p.
Catana, A. J., Jr. 1963. The wandering quarter
method of estimating population density.
Ecology 44:349-360.
Crockett, J, J. 1963. Preliminary studies in the
evaluation of grassland sampling techniques
in tall-grass prairie sites. Proc. Okla. Acad.
Sci. 43:43-46.
Dix, R L. 1961. An application of the pointcentered quarter method to the sampling of
grassland vegetation. J. Range Manage.
14:63-69.
Good, R. E., and N. F. Good. 1971. Vegetation
of a Minnesota prairie and a comparison of
methods. Amer. Mid. Nat. 85:228-231.
OF RANGE
MANAGEMENT
26(l),
January
1973
Johnston,
A. 1957. A comparison
of the line
interception,
vertical
point
quadrat,
and
loop methods
as used in measuring
basal
area of grassland
vegetation.
Can. J. Plant
Sci. 37:3542.
Levy, E. B., and E. A. Madden. 1933. The point
method
of positive analysis. N. Z. J. Agr.
46:267-279.
Morisita,
M. 1957. A new method
for the
estimation
of density by the spacing method
applicable
to non-randomly
distributed
populations.
Physiol.
and
Ecology
7:134-144.
Transl. published
by Nat. Res.
Council
of Canada,
Tech. Transl.
1167.
Ottawa, 1965.
Morris, M. 1962. In:Strickler,
G. S., and F. W.
Stearns. The determination
of plant density.
U. S. Dep. Agr. Misc. Pub. 940. 172 p. (p.
3040).
Parker, K. W. 1951. A method for measuring
trend in range condition
on national forest
ranges. Administrative
Studies, U. S. Forest
Serv., Washington,
D. C. 26 p. Processed.
Pielou, E. C. 1959. The use of point-to-plant
distances
in the study of the pattern
of
plant populations.
J. Ecol. 47:607-613.
Risser, P. G., and P. H. Zedler. 1968. An
evaluation
of the grassland quarter method,
Ecology
49: 1006-1009.
1956. Nonparametric
McGraw-Hill
Book Company,
York. 312 p.
Siegel,
S.
statistics.
Inc., New
Snedecor, G. W., and W. G. Cochran. 1967.
Statistical
methods.
Sixth Edition.
State Univ. Press, Ames. 539 p.
Iowa
Weaver, J. E., and F. E. Clements. 1929. Plant
ecology. McGraw-Hill
Book Company,
Inc.
520 p.
Whitman,
W. C., and E. I. Siggeirsson.
1954.
Comparison
of line interception
and point
contact
methods
in the analysis of mixed
grass range vegetation.
Ecology 35:431436.
A Comparison of Sampling Methods
in Dense Herbaceaous Pasture
PAULE S. POISSONET, JACQUES A. POISSONET, MICHEL P. GODRON,
GILBERT A. LONG
AND
Highlight: Several methods of vegetation sampling were compared in a very homogeneous herbaceous vegetation: Needle points, double metre points, bayonet points,
line transects, area measurements, and harvesting with sorting and weighing. Consistent
curvilinear relationships were found between species frequencies,
interceptions, and
biomasses obtained by several methods. These constant relationships allow the estimation of yields from the double metre points by simple and fast observations. The shape
of the curves suggests some ecological relationship between the vegetation attributes
analyzed.
Many problems arise in quantitative
sampling of dense herbaceous vegetations
because
of difficulties
in delineating
individuals or tillers and because of entanglement of species. The main problem
is optimalization
of time and resources to
attain sufficient accuracy. Another question is the kind of relations which exist
between estimates obtained by different
methods.
Many methods of measuring and surveying vegetation have been reviewed by
Brown (1954).
They differ
by the
attributes sampled, sampling techniques,
sample numbers, and types of lay-out.
Field observations may consist of species
lists, frequency (in sub-units), cover (total
or by species), density,
weight (by
species), etc. Sampling methods include
areas of various sizes and shapes, lines or
belt transects, and points. Point observations have been made by crosswire sighting tube (Winkworth and Goodall, 1962)
or with vertical, oblique, or horizontal
The authors
are plant ecologists
at Centre
d’Etudes
Phytosociologiques
et Ecologiques
(CEPE), Centre National de la Recherche Scientifique (CNRS), 34, Montpellier,
France.
They acknowledge
with thanks the guidance
provided
by Professor
Naphtali
Tadmor from
the Department
of Botany, Hebrew University
in discussing
and editing the
of Jerusalem,
paper.
Manuscript received June 27, 1970.
JOURNAL
OF
RANGE
MANAGEMENT
26(l),
point frame pins (Warren-Wilson, 1959;
1963). Parker (1951) and Long (1958)
have used lines and areas for locating
sampling
units.
Methods
have been
compared by HCdin and Lefebvre (195 l),
mountier
and Radcliffe
(1964),
and
others.
Optimalization
of sample numbers
with time has been discussed by Hyder et
al. (1965) and Fisser and Van Dyne
(1966). The distribution of samples may
be entirely subjective (i.e. as applied by
some
phytosociologists),
stratified
(Gounot,
1960), at random, or regular.
Numerous
combinations
at different
levels of these lay-outs have also been
used (Boudet and Bayens, 1963).
The present paper evaluates different
sampling methods by very detailed observations
on dense natural
herbaceous
pasture in Central France.
The Experimental
Area
The 0.4 hectare, very homogeneous
experimental plot is situated on a basaltic
plateau, West of Cantal in the Massif
Central of France, at 950 m altitude. The
climate is mesothermic
with a short
summer growing season. Precipitation
(rain and snow) totals 1600 mm annually,
and is regularly distributed all around the
year. The soil is deep, brown, mull moder
humus, very loamy (45%) and slightly
stony; pH is 5.5. The vegetation consists
January
1973
mainly
of perennial
grasses and forbs. The
three dominant
species are Festuca rubra,
Agrostis
vulgaris and Trifolium repens.
The area was grazed 6 months a year with
a stocking rate of l- 1.5 animal units/
hectare/year. For more details, see Daget
and Poissonet ( 1965).
Experimental
Procedure
Methods Compared.
The following sampling methods were
compared:
were defined
1) Needle points-Points
using a very thin needle, accurately
located by a special frame (Long et al.,
1972). The number of hits for each
species was recorded at each point. This
method gives the most accurate point
estimates but is time-consuming; it is used
as the “reference method” for further
comparisons. 8 x 160 points were located
at 2.5 cm intervals along 4 m lines.
2) Double metre points-Points
were
defined by vertical sighting on the edges
of a 2 m ruler (Daget and Poissonet,
1971). This method is much faster than
the previous one: 12 x 100 points were
recorded at 4 cm intervals.
3) Bayonet points-Points
were defined by the edge of the blade or a
“bayonet” (Poissonet et al., 1972), which
is convenient in tall herbage. Two hundred eighty-nine
points were located
randomly, 289 regularly, 256 regularly in
4 subplots of 64 1 x 1 m squares, or 0.5 x
0.5 m squares.
presence
4) Line transects-Species
was recorded in 25 cm segments of 4 lines
of 64 m each.
5) Areas with species ranking-Species
presence and their orders (De Vries and
De Boer, 1959) were done at regular
intervals of (a) “handfuls” (145 units)
which were harvested, (b) 0.5 x 0.5 m
squares (256), (c) 0.25 m2 squares (12),
65